ChatGPT can model FULL 3D SCENES now!? [Free Blender + ChatGPT Workflow]

[@OriBengal]

When is Netflix buying the Sci-Fi Cube series? I can't wait to binge that one!!!

[@danieldonaldson8634]

The basic mistake in all of this is to assume the language has the same form of precision as the visual domain. What actually happens is that in the interpretation of your prompts and the creation of output, chat GPT is effectively training you to accept its products as the basis for the next stage. at the core of the word imagination, is the word image. The point of creativity is that you form an image in advance of what you wish to create, and then apply skill that you have obtained to shape whatever medium you’re in to that image. What you measure during that process is the distance between the imaginative version and the physical version. But that isn’t what happens here. It’s basically the destruction of imagination in favor of mediocre substitutes that intercede before you get to the point where you realize your vision. When you look at the actual creative process, if you try to perform this prompt based process with someone else, or even yourself, the ambiguity in language would become clear. you give up and it would be clear quickly what the limits of language can actually do are.

[@skycladsquirrel]

We’re basically in the 8bit era of AI!

[@Bwaptz]

4:20 You made that cube look soooo good 🤣

[@badashphilosophy9533]

100%, i wanted that cube!

[@SenForte97]

I love how you actually explain how it works with the AI things and the knowledge in 3d you have, not like the instagram content creator that repetitively go "tHiS AI is cRaZy, this AI is gonNA tAke OvEr your job" etc with instant website and for only their engagement benefit.

[@neppallv]

This is super helpful, thanks for the clear explanation! I’ve been wanting to integrate ChatGPT with Python for a while, and your step-by-step guide made it easy to follow. Definitely going to try this out in my next project! Keep up the great work!

[@mickmack8026]

4:37 "mysteries within unfold" OMG so good...

[@adventuresinAI1982]

Writting complicated blender scripts and plugins is 95% of what I do with chatgpt. When I started a little over a year ago I knew nothing about coding and was basically just trying to create a few macros.

[@strikeforcealpha9343]

This blows my fucking mind.

[@SholvaBeats]

Impressive stuff, but nothing beats making everything yourself. It is an art form It is about creating something and getting better. No matter if it is digital or good old pencil drawings, i would never let a program take the fun away.

[@Zazuhtt]

You still need to have your vision anyways all it’s doing is helping you speed up the workflow

[@toapyandfriends]

Can you add the script here that you created during this tutorial that would be really great to get to download it.

[@Lancer95_305]

After I saw the donut I'm like u know what I better of doing it myself

[@paulhiggins5165]

I think using AI to generate 3D would be like using a teleporter to take a road trip- fast but somehow missing the point. The Techbro's seem determined to squeeze every last drop of joy out of the creative process, leaving us the dried out husk of a final result without the journey of getting there. Midjourney should really be called Endjourney- I'm surprised they missed this when they named it.

[@paulhiggins5165]

@@quercus3290 Why bother when soon an AI driven robot will come along and do it for me?

[@KarlJuluisGramota]

A bit different, like, for example, using "adobe after effects" in movies instead of a proper lighting system. It would look better than raw footage, but still kinda suck, but everyone will pretend that this was the whole idea from the begining. And descendants will spoke again in a commentaries about those "lost technologies of ancestors"

[@jvlbme]

OMG, are they actually _stopping_ you from doing it the dinosaur way?

[@paulhiggins5165]

@@jvlbme If you are working in a competitve enviroment where time is money then the answer to your question is yes- the 'dinosaur' way is no longer viable if AI can do it quicker and cheaper. But even for people who just do 3D for fun, a technology that could create instant results without any time or effort would not be that appealing I think. Sure they could try to pretend that technology didn't exist- but in reality I doubt many people will continue to do 3D for fun if AI could do it so much faster because they would feel stupid spending hours or days making something that an AI could create in few minutes- rightly or wrongly that's human nature- we are rational beings and as such are not really happy doing irrationl things. I can imagine some future version of blender where the UI is gone and replaced by a text box into which you can type your desired result- but I doubt many people would be using it- at least not many of the people who use blender today. Creativity is not just about the destination, it's about the journey- but with AI there is no journey, just a sterile process of prompt and iterate until you finally grow bored and call it done. Let me go all metaphysical on you and ask you to imagine what it would be like to be God- it would be great right? Everything you ever wanted could be yours instantly- in fact to a God wanting something and having that thing would be the same thing- literally instant gratification, how amazing would that be? The moment you even think of something you want- it's yours! Zero effort. But here's the snag- epic, eternal, never ending boredom- the moment you can have everything nothing is worth having. This is the long term promise of AI- to create a scenario in which humans can have anything they want, instantly and without effort. The question is what do we call this scenario- is it a paradise on earth? Or is it a very special kind of Hell?

[@jvlbme]

@@paulhiggins5165 I know where you are coming from, but the thing is 'just' prompting (and it will mostly be by voice, btw, not a textbox, and probably not within Blender) IS creative. You are going to want to _change_ the results, try different colors and textures, try another composition, another solution. It's MORE creative even.

[@snoomr]

omg the cube music clip

[@CodeLaniakea-tf5no]

I believe he using Suno AI

[@CodeLaniakea-tf5no]

cool! Thanks to ChatGodPT it will learn more and more about scripting for Blender. This way, it can provide us with more complex mathematics for intricate objects, like trees with branches, leaves, logs, textures, lighting, shading, coloring, and more.

[@maddercat]

oh shit so amazing and funny. the cube song had me rolling.

[@Dohpamine_Design]

i passed out haha

[@iresolvers]

this crazy cool can't wait to see where this goes in the future!

[@MafutaGames]

That Sci-Fi Cube Rock! Going Viral Soon! Interesting future ahead.

[@playthisnote]

As it’s been suggested let’s use vision to evaluate the objects it creates. I’m sure we can find a version of blender or something that can be put into this type of loop. I’ll look into it as well

[@tabathacarla]

I cracked up at the sci-fi cube

[@lucidbonez5669]

i think its important to have a healthy balanced workflow between utilizing the ai and having your hand in the mix so to speak, so you remember that its a tool and you are the director, and so you can finely tune it to your liking with the power of ai at your disposal

[@TheiLame]

Hi, what rendering workflow was done to generate the scene using blender at 7:58?

[@halvsketchy9293]

Was thinking the same thing lmao. Looked like the work of people who fix people's drawing and say "then some finishing touches and we're done"

[@SeaHorseOo]

Playing the One director with an AI assistant in Blender is interessting enough. It would inevitably need to be able to see sketchs and read visual instructions like rough storyboarding.

[@strikeforcealpha9343]

Question, can this be done on Unreal 5?

[@strikeforcealpha9343]

This is mindblowing oO

[@spinx2273]

LOL Sci-Fi Cube is a banger!

[@YesRelax]

Do you delete every code when you enter a new code?

[@muneerty]

Only if there's was a way to make gpt communicate with itself by taking screenshots of the scenes it created and letting it give feedbacks, and thus it makes changes to the codes based on its feedbacks.

[@playthisnote]

There is

[@hamidthephysicist6376]

ROLF` the sci-fi cube killed me` . Thank you for your amazing content

[@leretah]

Excelente contenido, muchas gracias por el tiempo que te tomas en cada tutorial.

[@SUPERPIXELPLANET]

This was such a fun video! I think it would be awesome if you started making videos that show off some of the experimentation you do like this. Not every video needs to be your amazing workflows; some of them can be some of the things you’re doing to find your next amazing workflows. And we’d get uploads more often :D haha

[@mickmumpitz]

Thank you so much! That's my plan. :) I have so many fun and exciting experiments that I never get to show because these big videos take up so much time. I definitely plan to do more stuff like this in the future!

[@SUPERPIXELPLANET]

@@mickmumpitz That is awesome, man! I really appreciate what you do! I’ve learned and lot from your videos 🙏🏼

[@ItsWesSmithYo]

Main problem I’m having is ChatGPT is current to 2.92 but can’t really do much except make guesses about feature that are new since then. Anyone else figure out how to get around that?

[@Tio6835]

Can you make one comfyui tutorial with the screen share node?

[@delabeccamarie4612]

Okay but how did you make the music that’s what we actually wanna know

[@Visionnaire__Studio]

Awesome video! If one had to start from zero with 3d animation, where would you recommend to start from and what tools would you recommend?

[@mickmumpitz]

Thank you! Definitively start with blender. It’s a completely free 3D software with amazing functionality and a lot of tutorials on YT. I recommend you start with blendergurus donut tutorial I mentioned in the video. SourhernShotty also has some fun tutorials for beginners.

[@Visionnaire__Studio]

@@mickmumpitz Amazing advice, thank you so much really!!

[@jeffsmith9384]

Hey, a glowing blue sci-fi cube worked for Marvel, they called it the tesseract

[@PulpetMaster]

have you tried any node or local llm's like llama (codellama) mistral or any other option?

[@Polyroad3D]

OMG AI is getting closer and closer.

[@AbsolutelyForward]

😂 your content is just pure joy - and also very helpful! Thanks

[@fhfdgddggdg]

can i made 20th fox logo snow in blender gpt4 this video

[@oraclecamp]

So it cannot Sculpt via the Python Script? I remember not all Blender functions are available on the Console.

[@electronicmusicartcollective]

Gude, Deine Videos feiere ich schon so lange. TOP, weiter so. Auch immer einen guter Vibe dabei. Mercie... btw schöne Referenz zu Andrew Price, welchen ich auch von Anfang an verfolgt und er ist für einen großen Teil meines blender Wissens verantwortlich :)

[@GabrieloTomorrow]

You know what? This is very exciting! Sure, it's currently very limited and it looks kinda crappy... But AI is moving incredibly fast, and if the ChatGPT that we currently have can already do something like this, then I bet that it won't take more than 2 years until an AI model can actually make truly impressive scenes.

[@gabrielsandstedt]

Try this again with claude sonnet 3.5 it is so much better at scripting :)

[@rameshdevasi6720]

how to texture a 3d model from all side using AI, now i generate image for front and side angle, but its not seamless.

[@cupotko]

I need the complete Sci-Fi Cube song and lyrics!

[@Davidgetz1]

The real problem here is optimization for example: if you do an scene shader, etc and how long it takes? and with chagpt if we make really dificult prompts and mixing both how long it would takes?

[@altamiradorable]

For longer prompts use o1 not gpt 4o

[@cartoonbay]

can you please teach us how to fix 3D models from AI 3D generators? I work with Unreal engine and never used Blender but sometimes need to fix FBX models that I generated in Hugging face like filling holes and cutting extra hands ....

[@badashphilosophy9533]

could you train chatgpt or other models to be pretty exceptional at this though? and even train it potentially on its contextual limitations so it breaks its own outputs into managable pieces perhaps

[@pm1234]

Realizing we prompt like directors with a vague idea of what we want! That's how they see us! 😅😄😏

[@Glowbox3D]

I think it's pretty cool that GPT basically created my logo when asked to do what it wanted. See, I knew I was onto something! lol. Great video. This stuff is getting crazy. It looks like it's creating old VRML worlds.

[@richoffks]

Bro did a donut and I went straight to a complex human super hero character 😭😭 I be breaking ai, it’s like “this is too complex” smh 🤦🏾‍♂️

[@SufyanBloch-vy7ps]

Please make a video on this topic that how we can create our ai ed avatar and animate it with the help of our mobile camera. Like in your video in video

[@s1amvwbug]

Question: I think if you would have given a picture and explain what you wanted to ChatGPT. I believe the output for Blender would of been better.

[@phu320]

5:00 LOL

[@nholmes86]

yes in the future ai will do 3d, and you just prove it!

[@reezlaw]

I think we're minutes away from a voice controlled Blender. All it takes is an addon you can talk to and that interfaces with Blender's scripting or some other way, in realtime. Make me a house... no, make it two floors... that's right now add a garage to the right... make it wider... ok now add a road ETC

[@elvenprints]

I feel relieved there is still something to do for humans. 😀

[@FelipeSotoSCel]

4:17 hahahaha

[@3DMVR]

I need an ai for suggestions in the geonodes

[@jvlbme]

No, we won't be directors typing prompts, we'll _say_ them. 😁

[@METTI1986LA]

Why take something easy like a donut and turn it into the most complicated thing ever…

[@KarlJuluisGramota]

Cause they have a new technology and they need to use it disregards the common sense.

[@reezlaw]

It's the concept. Right now it's limited and apparently useless but at some point it will become more convenient than making it by hand

[@KarlJuluisGramota]

@@reezlaw IF it become more convenient. The optimization problem is still here.

[@gumvue.studio]

yes, this is very good. but chatGPT yet have place to grow

[@peteriliev]

The real question is how far (deep) we (they) will go…

[@elarcadenoah9000]

gracias por tus videos

[@High-Tech-Geek]

7:56 OMG, how?

[@mickmumpitz]

Check out my last video for the full workflow!

[@alexmehler6765]

answer is no ... save your time

[@cihatfg]

nice

[@mrallen1971]

that item at 3:22 looks like mince-meat frosting with maggot sprinkles. gross.

[@cyrusthegreat184]

"I dont wanna learn to do anything i just want to make money off it"

[@MoonGluedStudios]

Anybody getting results from this they care to share?

[@NwoRun]

what a fun episode.

[@KarlJuluisGramota]

There one problem in making ChatGPT work for you or instead of you, that everyone forget - chatGPT is not a person (in basic words - it is a robot and not a legal entity) and it is a tool which works online, and belongs to a "third party" of it creators. What about code it make and inellectual property on that? Who owns what? Could it be, that this program has a derrective that would sort tasks, take those that are interesting to its owners and keeps the better solution to them, while drop half baked and unoptimizaed garbage to its real author?

[@Blocky007]

Wow... :). Btw... what works well in Webdesign with ChatGPT might work well with Blender: Take screenshots of the product the code produced and feed it back to ChatGPT. So that it can actually SEE what it did. NOW you describe what you want to have improved. Tends to create better results! Would love to see you try :)

[@LouisGedo]

👋

[@pixel-ink]

This is why I don't use Ai to create Blender objects or scenes. I rather use my own creative skills. Ai robs you of being a true creator.

[@ThanosDestroyeryearsago]

Okay

[@badashphilosophy9533]

funny lol

[@juniortech8832]

The chatgpt created buildings, looked like a computer motherboard😅

[@marilynlucas5128]

I just don't understand why so far geniuses have not been able to make AI model accurately. It's such an open field for anyone to capture. Manual modeling should be dead by now. SHAP-E is already available so what's the hold up? I would have focused on developing a full AI system for modeling but I'm just too busy now and I hope someone else will figure it out. Mamba is a state space llm architecture that can be used to guide modeling in blender and other 3d systems. It's about time we move away from manual modeling. I want to be able to accurately model entire production lines and factories

[@SholvaBeats]

Maybe learn how to model ?

[@marilynlucas5128]

@@SholvaBeats Interesting suggestion but I've figured in the not so distant future, AI will be able to model. An llm like Mamba with it's state space architecture can think in geometric terms so it's only a matter of time before LLm's can model anything! This is why I don't want to waste the time to learn modeling

[@KarlJuluisGramota]

Then you will discover another problem all "those video with happy smiles" never talk about - the hardware limitation.

[@blacxthornE]

ok i don't care about chatgpt at all but i saw the thumbnail and it annoys me. you don't pixelate an image and then rotate it. you rotate it first and then pixelate it. your thumbnail shows a pixelated image with diamond-shaped pixels. disgusting.

[@johnylitalo4163]

Not interested.

[@boxer8657992]

Not impressed

[@newb7794]

you're just bad at prompting

[@ItsWesSmithYo]

Equal problem is ChatGPT knowledge base is limited, so….😂

[@TigeyPuss1]

I'm not an artist. I can make rudimentary things like a circle or square, but I can only do a little more. So, I hope I can get ChatGPT and Blender to take my ideas and make them outstanding. I want to show two amino acids coming together to make a dipeptide. The -OH on one amino acid is combined with the H+ ion from the other amino acid, which produces a water molecule. This is when the two amino acids join and become a dipeptide in a process called condensation. That could be more exciting. However, this basic process can produce proteins, enzymes and such, that do wonderful things inside a living cell. An important point in this process is that the joined amino acids are more likely to split apart than join together. This means that solutions of amino acids can't form proteins. This has significant philosophical implications. I want to visually show the process and why dipeptides don't continue to grow into tripeptides and eventually become functional proteins. Is it possible to show this by animation?

[@ozgureren8925]